US consumer price index, all items · United States · 3 months ahead

2026-10: the model assigns an 80 percent probability to 333.8 to 336.6 index 1982-1984=100

Point estimate 335.7 index 1982-1984=100 from Drift, against 332.8 index 1982-1984=100 observed in 2026-07. drift beats last_value on out-of-sample MAE by 63.9 percent over 44 cutoffs (minimum 24, margin 5 percent).

ModeledMetric page

Distribution over the history

Fan chart

Solid: observed (index 1982-1984=100). Dashed and shaded: model point estimates and 80 percent intervals from empirical residual quantiles. Filled dot: the record on this page. Green or red dot: the realised value, inside or outside its interval.

Quantiles: p5 333.5 · p10 333.8 · p25 334.6 · p50 335.2 · p75 336 · p90 336.6 · p95 337.6
Source: empirical quantiles of 35 in-window 3-step residuals, in log space; no normality assumed

The record

Immutable fields

Key: us-cpi-all-items:USA:3
Id: 1edc7e36-bf33-4734-92c7-74d48fd235a8
Created: 2026-09-07 22:16 UTC
Data cutoff: 2026-09-07 22:16 UTC
Origin: 2026-07 = 332.8 (obs 167372)
Target period: 2026-10 (92 days)
Point: 335.7 index 1982-1984=100
Interval: 333.8 to 336.6 at 80 percent
Model: drift models_v1.targets_v1
Baseline: last_value
Snapshot: 3eb2681555281150bff8d4f24e39c0921d910209e6f88ad7af826b6e13542199
Data mode: ingested_at
Status: active

Drivers

Statistical association, not causal

Drift uses only the target's own history; there are no driver contributions.

Outcome

Not yet realised

The observation for 2026-10 has not been ingested. The weekly forecast.evaluate job attaches it when it arrives; the record above does not change.

Scenario assumptions

What the forecast takes for granted

No structural break in us-cpi between the data cutoff and the target period.
The source keeps publishing the series on the same definition; a rebasing or redefinition invalidates the comparison.
The average historical change per period continues.

Falsifiers

What would show this forecast wrong

If the first published value of us-cpi (USA) for 2026-10 is below 333.8 index 1982-1984=100 or above 336.6 index 1982-1984=100, the 80 percent interval is falsified.
The model assigns a 10 percent probability to a value below 333.8 index 1982-1984=100 and a 10 percent probability to a value above 336.6 index 1982-1984=100; interval misses should occur about one time in five over many forecasts, and a run of misses well above that rate falsifies the calibration.
The point forecast is above the last observed value (332.8 index 1982-1984=100); a realised value at or below that level falsifies the direction call.

Competing models

Scorecard for this target and horizon

Model	Status	Cutoffs	MAE	RMSE	Direction	80% coverage	Skill vs last value
Drift models_v1.targets_v1	champion	44	0.8919	1.2	100%	68%	63.9%
Holt (exponential smoothing with additive trend) models_v1.targets_v1	challenger	44	1.28	1.62	100%	77%	48.2%
Last value models_v1.targets_v1	baseline	44	2.47	2.73	0%	93%	0.0%
Linear trend models_v1.targets_v1	challenger	44	4.26	5.02	91%	57%	-72.7%
Ridge regression on lags and drivers models_v1.targets_v1	challenger	44	0.906	1.25	100%	66%	63.3%
Seasonal naive models_v1.targets_v1	challenger	44	10.69	11.21	0%	98%	-333.3%

Model card

Drift

Purpose. Point and 80 percent interval forecasts of us-cpi (USA) 1, 3, 6, 12 months ahead. Owner: Forecast Lab (HV 3.0).

Method

The forecast extends the mean first difference of the whole training window from the last value: last value plus horizon times the average change. The series is modelled in natural logarithms and transformed back, so intervals are asymmetric in level terms.

Uncertainty

Intervals are empirical: the model is refitted at earlier origins inside the training window, the realised errors at the same horizon are collected, and the interval is the 10th to 90th percentile of those errors around the point forecast. No normality is assumed.

Training and evaluation

Training window: 954 periods, 1947-01-01 to 2026-07-01. Evaluation: rolling-origin backtest over 44 cutoffs using pseudo real-time data (the earliest ingested value per period; no vintage archive exists for this series, so revisions between the cutoff and today are not reproduced).

Known limitations

Structural breaks after the cutoff are not modelled.
Driver contributions are statistical associations, not causal effects.
Promotion to champion requires beating the last-value baseline on out-of-sample MAE by at least 5 percent over the minimum number of cutoffs; otherwise the baseline stays champion.
Retraining: every scheduled forecast.run refits on the data then available; backtests re-run from /admin/forecasts.

Version chain

Supersedes

This is the first record for its key.